"Unlimited Detail" A technological revolution in animation?

I read a Game Informer article the other day about this new technology that, if it does all the creator says it can, could possibly change the game of animation and graphics. Here's the GI article.

Rather than use polygons--which are essentially flat pictures put together to look 3D--this system imitates how the real world works by building things out of "atoms" in the virtual world. This allows for incredible detail and scalability. If you don't want to read the long article, here is Euclideon's YouTube page.

It's interesting, in the article the guy says this became possible because he didn't get a typical university degree, so he wasn't taught the "proper" way to think like a computer scientist. When he saw a problem, he didn't look for a typical/easy solution, he dug deep into the problem and built a solution from the ground up. This is an insane adventure in programming. I'd love to see how it works on a technical level.

I'm not a believer.

I'm pretty annoyed really. Its just another voxel engine.

Even if he did come up with a new voxel-picking strategy to determine what gets drawn, voxels still have some major shortcomings.

1. Animation is nigh-impossible.

Consider an object built of millimeter scaled voxels. How would you rotate them to know their positions in world-space for drawing? That's hundreds of thousands to millions of point rotations. Even if an uber-effecient method of rotating points came into play, thats still too much calculation for a game world filled with loads of animated objects. Current games that use voxels revert to polygonal models for the animated stuff. If this guy did that his claims of software rendering would fall through.

You can see in the UD videos that each of the tree models are the same, and that every repeated object is in the same orientation. He can't rotate them.

2. Memory concerns.

For totally unique worlds you have to be able to save each of those points and their colors.

https://www.youtube.com/watch?v=_CCZIBDt1uM&feature=channel_video_title

That's a voxel engine. The guy can supposedly compress each voxel point down to around a byte per voxel, and yet that little map chunk is still 500mb. Makes you wonder what kind of screwed up file-size WoW could reach if it were a voxel game.

Also you can notice in the unlimited detail videos the unattractive amount of repetition. This is no doubt to save on memory, but it does hurt the 'unlimited detail' argument. You cant go carve your name into a wall because it would have to re-save the chunk of affected land and take twice the memory.

The other argument is converting polygon models to voxels at run-time. This is fine and all, but it does make loading times longer and hurts your ability to stream models from the hard-drive in real time. It also hurts your ability to detail objects at smaller scales. The smaller you make the triangles of a model, the closer you get to voxel rendering. If you throw tons of triangles at a model, you might as well have just saved it as voxels in the first place. If you don't throw tons of triangles at it, you might as well not be using a voxel engine because you wont get the full detail out of a model. Meaningful games will have to pick a balance between detail and memory.

3. It makes lighting hard.

The nice thing about triangles is that they have normal values intrisically built into them. You can readily grab a surface direction for lighting. Voxels make that a bit less trivial. You could get a normal vector by building a surface from a voxels surrounding 26 neighbors and extracting it from that, or ray-casting from the point in question to every light that affects the voxel to check for geometry collisions. Its a lot of wasted effort. Lighting is already the most expensive part of a rendering pipeline, so the extra lighting computation makes the prospect of 'interesting' voxel lighting pretty impossible with current technology.

I would never make a game with lighting like that though. Lighting and shadows are too important for a game's setting to take a step back like that. Think the sneaky shadow crawling of splinter cell, or the shadowy corridors of doom/quake. Without nice lighting, the settings for those games would have suffered.

Also, the reflection of the water in the UD videos shouldn't be considered lighting. They're pulling the cheap trick where they mirror the geometry of the scene above the water. They're drawing the same scene twice, and throwing a water tinge on the stuff under the water.

It's interesting, in the article the guy says this became possible because he didn't get a typical university degree

Arguably this shouldn't matter. The people who are into graphics research have more than likely gotten that way out of their own personal interest and fiddling than from their classes. Classes are only good for "keyword wetting" and imposing deadlines. They tell you what to google when you're clueless, and give you a deadline to get your stuff together.

tldr; That guy is a hack. >_>

It's been a while since I read it, but I'm pretty sure he addressed all of that.

For the repitition, he said they had a limited time to put the demo together and no real artist to work on it. The lack of variety isn't because of limitations in the technology, they just didn't scan in enough real world objects.

As for other things like lighting, animation, and collision detection, he says they are in the process of being developed and will be in the final version. I guess it doesn't really matter how we feel about it; we'll see in a year or two if they come out with a working technology.